This PDF is a selection from an out-of-print volume from the National Bureau
of Economic Research
Volume Title: American Economic Growth and Standards of Living before
the Civil War
Volume Author/Editor: Robert E. Gallman and John Joseph Wallis, editors
Volume Publisher: University of Chicago Press
Volume ISBN: 0-226-27945-6
Volume URL: http://www.nber.org/books/gall92-1
Conference Date: July 20-22, 1990
Publication Date: January 1992
Chapter Title: American Economic Growth before the Civil War: The Testimony
of the Capital Stock Estimates
Chapter Author: Robert E. Gallman
Chapter URL: http://www.nber.org/chapters/c8008
Chapter pages in book: (p. 79 - 120)
2
American Economic Growth
before the Civil War: The
Testimony of the Capital
Stock Estimates
Robert E. Gallman
2.1 Introduction
Robert Giffen, of paradox fame, thought estimates of aggregate wealth have
eight uses; the following have immediate relevance:
1 . To measure the accumulation of capital in communities at intervals of
some length . . .
2. To compare the income of a community, where estimates of income exist,
with its property . . .
4. To measure, in conjunction with other factors, such as aggregate income,
revenue, and population, the relative strength and resources of different
communities.
5. To indicate generally the proportions of the different descriptions of property in a country to the total-how the wealth of a community is composed.
6. To measure the progress of a community from period to period, or the
relative progress of two or more communities, in conjunction with the
facts as to progress in income, population, and the like; to apply, in fact,
historically and in conjunction with No. 1, the measures used under the
above heads 2, 3, 4 and 5 for a comparison at a given moment. (1889,
136-37)
Robert E. Gallman is Kenan Professor of Economics and History at the University of North
Carolina, Chapel Hill, and research associate of the National Bureau of Economic Research.
The research underlying this paper was funded by the National Bureau of Economic Research,
during my tenure as an Olin Fellow, and by the National Science Foundation, to which organizations I express my gratitude. In another form the paper was given to seminars at the California
Institute of Technology, the University of Chicago, the University of California at Davis and Los
Angeles, and Northwestern University. At all of these seminars I received useful suggestions,
especially from Lance Davis, David Galenson, Morgan Kousser, Kenneth Sokoloff, and Sokoloffs graduate class in economic history. The discussant of the paper, Stanley Engerman, was, as
always, most helpful.
80
Robert E. Gallman
Simon Kuznets, who made use of the list, said that the distributions alluded
to in item 5 should include the size distribution of wealth (Giffen 1889; Kuznets 1958; see Engerman and Gallman 1983 for more on these issues).
The remarks of Giffen and Kuznets provide a justification for this paper, a
list of things to include in it, and a set of suggestions as to how it should be
related to the other papers prepared for this conference. Particularly attractive
is Giffen’s notion that many different types of aggregate series, as well as
compositional indexes, should figure in the measurement of growth. He
would have felt comfortable at the conference at which this paper was given,
as the participants approached the questions of economic development and
standards of living from various directions, using data on labor force, income,
wealth, consumption, wages and prices, productivity, and heights.
Kuznets believed that a perfectly realized index of development would trace
out shifts in human material welfare. Such an index could be employed to
measure changes in the standard of living-so long as we understand that
term to refer to the material aspects of life-without the need to introduce
other measurements. But Kuznets was well aware that the indexes of development with which scholars must work are far from ideal and that, therefore,
a variety of them may be required. In the spirit of Giffen and Kuznets, then,
this paper treats the capital stock series as one that bears on the standard of
living, rather than as one that measures it.’
Capital stock series have two possible conceptual relationships to economic
growth (Gallman 1986). First, such a series may be used to measure the
wealth accumulated by a society. The accumulation will be influenced by the
economic performance of the economy in the past, by the degree of frugality
displayed by its people individually, by the success the society had in its military activities, and by communal saving and investment decisions. The measure is clearly different from income, in that it relates to a stock collected over
a period of years, not a flow during one year. Income and capital series are
likely to change at different rates, then, at least in the short run. But the two
types of series do both bear on the material well-being of the people of the
society.2In the very long run they are also likely to exhibit roughly the same
1. This is not the first attempt to study the American economy before the Civil War by examining capital stock data. See, for example, Jones (1980); Goldsmith (1952, 1985); Davis, Easterlin,
Parker et al. (1972). I think, however, that it is the most serious effort to assure that the various
estimates are consistent from one date to the next.
2. One virtue of a capital stock series as an indicator of growth is that the short-term movements
of such a series are likely to be much less violent than, for example, the short-term movements of
a true income series. If estimates are available only at intermittent years, the rates of growth
computed from the former are much less likely to be influenced by transient phenomena than are
the rates of growth of the latter. It should be said, however, that this distinction probably does not
apply to the income estimates for the years before 1840 that were put together by Thomas Weiss
for this volume, since his estimating procedure does not pick up the effects of short-term influences on income, nor is it intended to. Weiss’s estimates come close to describing the output
capacity of the economy, rather than actual output.
81
American Economic Growth before the Civil War
growth rates, so that a capital series can serve in some instances as a proxy for
income.
The second approach is to view capital in its capacity as an input into the
production process. Whereas the first approach looks chiefly to the past and
sees capital as the accumulation created by society, the second looks to the
present and the future. It sees the capital stock as one factor influencing current production, as well as production to be expected of the economy in future. Clearly, such a series is particularly useful when combined with estimates of the other inputs.
If the direct relationship between real capital and material well-being is to
be examined, the capital stock series should be deflated by a consumers’ price
index. That is, the stock should be appraised in terms of its equivalent in
consumer goods. If, on the other hand, one is concerned with productive potential, proper deflation is in terms of the prices of the components of the
capital stock. Both forms of deflation are employed in this paper. That is, the
capital stock is treated as an index of both the material well-being of the society and its productive power.
The concept of capital is elastic. Some analysts have included land and
investment in humans as elements of the stock. For most purposes, it is best
to treat land as land and human capital as a characteristic of labor. In the
present instance, the second preference makes a virtue of a necessity: there
are no comprehensive estimates of human capital covering the full period of
interest here. This paper introduces a set of estimates of the land stock, but
they are not treated as part of the supply of capital.3
Although land is not included in the capital stock series of this paper, improvements to land are. In this respect the series is unconventional. Most capital estimates include structures but omit other important improvements, such
as the clearing and first breaking of land. In this paper a conventional series is
presented and is linked with estimates extending well into the twentieth century, for comparative purposes. But the series that is subjected to the most
intense examination is one that includes the value of land clearing and breaking. These activities took up a substantial part of the work time of agricultural
workers and made an immensely important contribution to the capital stock
before the Civil War. They cannot properly be ignored.
3. Should the value of slaves be counted as part of the value of the capital stock? If we are
interested, say, in the savings and investment behavior of planters, then the answer is surely yes.
This paper is not concerned with that topic. It is concerned with the measurement of long-term
economic growth. Slaves are regarded as part of the labor force. They are also treated as part of
population, for purposes of computing per capita levels of the capital stock.
While I will present no estimates of the value of human capital, the general pattern of change in
this variable across the period under review here is quite clear. Both the fraction of the population
of children attending school and the length of the school year increased as time passed, as did the
fraction of the work force holding semiskilled and skilled jobs. The rate of increase of human
capital is therefore almost certain to have risen as time passed. See Fishlow (1966a, 1966b) and
Uselding (1971).
82
Robert E. Gallman
The value of consumer durables is also sometimes incorporated in capital
stock estimates, but appears in only one table in this paper, because appropriate figures are only intermittently available. The loss is not great. The value
of consumer durables was small, compared with the rest of the capital stock,
through most of the period considered in this paper, and the rate of change of
the capital stock is approximately the same, regardless of whether or not durables are treated as capital (see table 2.2).
The United States (for convenience, the term will be applied to the colonies
of 1774) began life as a debtor nation and gradually shifted to the position of
a creditor nation. Ignoring recent experience, the national capital stockwhich measures the net capital holdings of Americans-grew very much
more rapidly over time than did the domestic capital stock, defined as capital
physically located in the United States, regardless of who owned it. Through
the rest of the paper, both series will be examined, although most attention
will be devoted to the domestic capital stock.
The title refers to the period before the Civil War, but the series introduced
will typically cover a much longer stretch of time. The fundamental questions
at issue have to do with persistent changes, and these questions can be properly addressed only if data bearing on long periods are available. Some rates
Table 2.1
Capital and Wealth, 1774 and 1805, Estimates of Jones, Goldsmith,
and Gallman (millions of current dollars)
1774
Jones
All structures
All land improvements
All privately owned real
estate
Shipping
Other producers’ durables
Inventories
Animals
Total domestic capital
International claims
Total national capital
Total domestic capital,
including the value of
clearing and first
breaking of farmland
Total private domestic
capital, plus land
1805
Gallman
Goldsmith
Gallman
370
352
732
180
250
8
13’
15
20
42
39
42
40
32
100
60
602
- 80
522
80
65
336
160
993
- 57
936
284
327
Sources: Jones (1980, 90,converted to dollars by means of the exchange rate on page 10);
Gallman, see text; Goldsmith (1952, 315).
aIncludeshousehold equipment.
83
American Economic Growth before the Civil War
of growth covering relatively short intervals-such as the two decades 184060,which have been the focus of much scholarly interest-will be exhibited,
but the reader should bear in mind that such rates are influenced by short-term
phenomena. They cannot be used as the exclusive means of identifying shifts
in trend rates of growth.
Section 2.2 deals briefly with the nature of the data underlying the estimates
and the broad rules guiding the estimating procedure. Section 2.3 treats the
rates of growth of the real capital stock and the real capital stock per capita,
with the purpose of putting growth before the Civil War into historical perspective. Giffen’s suggestion that the rates of change of capital and income be
compared is taken up.
Economic development involves structural shifts as well as growth in the
aggregates. Section 2.4 treats the changing composition of the capital stock
and shows its connection to the nature of American economic development.
Section 2.5 brings together estimates of all three factor inputs and combines
them into several series describing the growth of total factor inputs. Estimates
of changes in total factor productivity are presented. Section 2.6 is a summary
of conclusions.
The data on which the estimates rest pose many problems. Appendix A
takes up several important features of the series and considers a few tests of
the most affected components.
2.2
The Estimating Procedures
Estimates were made for the years 1774, 1799, 1805, 1815, and 18401900 (at decade intervals) and for various dates in the twentieth century. As
Giffen points out, capital series can be used to study economic growth, “regard being always had to the fact that the data and methods employed are
sufficiently alike for the special purpose in hand” (1889, 136). The object of
this section is to consider whether “the data and methods employed are sufficiently alike for the special purpose at hand.” The subject is treated further in
appendix A.
The current price capital stock estimates for 1850 and 1860 are based
chiefly on census materials, which have been tested in a variety of ways and
adjusted to make them consistent from one date to the next and to make them
conform to an appropriate concept. The best overall tests that have been conducted so far are checks against perpetual inventory estimates derived from
measurements of investment flows. The results of the checks are excellent
(Gallman 1986, 1987).
The 1840 figures were similarly derived from census data, augmented in
various ways, chiefly by contemporary estimates produced by Ezra Seaman.
The census in 1840 was quite different from the ones that followed. It was
administered under a different law and asked different questions. For some
purposes it is quite good, but it is clearly weaker than the later censuses as a
84
Robert E. Gallman
source of material for the estimation of the capital stock, although it has survived testing quite well.
The estimates for these three years and some of the tests that have been run
are described in Gallman (1986, 1987). These sources also contain the estimates for the years 1870-1900, which will be used in the present paper to put
the experience of the years 1774-1860 in context. Tests of the post-Civil War
data by means of perpetual inventory estimates suggest that the 1870, 1880,
and 1890 aggregate capital stock figures are unlikely to be perfectly consistent. It appears probable that the 1880 figure is too low. Calculations of the
rates of change of the capital stock from this series are therefore likely to
understate the true rate of growth for the period 1870 to 1880 and overstate it
in the years 1880 to 1890, matters of no great importance for present purposes.
The twentieth-century figures were assembled by splicing the nineteenthcentury estimates to Raymond Goldsmith’s series, which are based on perpetual inventory procedures (Goldsmith 1982). As indicated above, census-style
and perpetual inventory estimates appear to be roughly comparable.
The estimates for the years before 1840 come from a variety of sources
quite different from the censuses, which increases the risk that the capital
stock estimates based on them may not be consistent, one with the other, and
all with the figures for the years 1840 onward. The data that are farthest removed in type from census data are the ones underlying the capital stock figure for 1774. These data were taken chiefly from Alice Jones’s (1980) work
with probate records. The figures for 1799, 1805, and 1815 rest principally
on sources that are more likely to be consistent with census records: the direct
taxes of 1799, 1813, and 1815. (I used the data in Blodget [1806] 1964, Pitkin
1835, and Soltow 1984.) The 1805 estimate is based on the work of Samuel
Blodget ([ 18061 1964) and Raymond Goldsmith’s (1952) adjustments
of Blodget’s work. The principal underlying source is the direct tax of 1799.
Blodget apparently carried the 1799 data forward to 1805 at a rate of growth
he believed most probable. The 1805 estimate falls out of line with those of
1799 and 1815 and is probably too high. The history of the period leads one
to expect a higher rate of growth between 1799 and 1805 than between 1805
and 1815, of course, but not quite so high as the Blodget data suggest. Of
course it is possible that the 1805 figure is close to the truth and that the other
two are too low, but I do not think that is the case. It is also possible that the
bias was introduced by my adjustment of the Blodget data (see table 2. l), but
I doubt that is so.
The 1774 through 1815 estimates depend on the sources listed above, augmented and adjusted so that the same concept of capital underlies each final
aggregate figure, and so that the same estimating principles are applied in each
case. The last point is an important one. While accurate estimates were sought
in each instance, it seemed clear that it would be better to have a series for
85
American Economic Growth before the Civil War
which the general level might be wrong, but which describes the rate of
growth in a reasonably accurate way, than to have one for which the individual
estimates might be closer to the truth, but which gives a more strongly biased
account of the rate of growth. The choice made was always for consistency
rather than for perfect a c c ~ r a c y . ~
Table 2.1 compares some of the details of the new estimates with those
provided by Jones and Goldsmith. As will be evident, the adjustments made
to the Jones figures were relatively unimportant, so that the new estimates tell
very much the same story as do the data taken from Jones. The differences
between my estimates and Goldsmith’s are greater, and are particularly pronounced with respect to inventories of all kinds. Goldsmith’s estimates seem
too low to me; for example, imports in 1805 ran around $150 million, and
imports represented a relatively small part of total economic activity, even in
1805. Even a very modest estimate of the fraction of imports held, on average,
in inventory across the year would leave very little for inventories of domestic
goods, were we to accept Goldsmith’s figure for total inventories. But the
question of the appropriate level of inventories in 1805 is perhaps not the
important issue. The important point is the one made in the previous paragraph. In building the inventory estimates for all of the years, 1774-1900, I
have tried to follow consistent methods and have paid more attention to consistency than to the specific level of any one estimate. Consistency permits
appropriate comparisons to be made across time, an important desideratum.
Users of capital stock series for the nineteenth century, then, would be well
advised to use either Goldsmith’s estimates or mine, but not some combination of the two.
All of the capital figures are expressed in market prices or in net reproduc4. For example, imagine a series that has true values of 100 and 200 in two widely separated
years. If the estimatesproduced for these years are each too large by 10percent, then the estimated
series and the true series will describe the same rate of growth. That would not be the case if the
estimates were closer to the truth in each year, but deviated from it by different percentages-say
the first estimate amounted to 95 and the second to 210. Obviously, one cannot know with certainty that the first circumstance or the second holds in any given instance. But there are cases in
which one has the choice of following a consistent procedure and using consistent data from one
date to the next, in full knowledge that the results are unlikely to be exactly correct, or employing
different methods and bodies of data, in an effort to come as close as possible in each year to the
true value. Where I was presented with these options, I chose consistency. (Emerson, after all,
only deplored a foolish consistency.) But consistency at the component level does not guarantee
unbiased rates of growth at the aggregate level. Suppose that the level of each component series is
biased by a given percentage in each year, but the given percentage varies from one component to
the next, by amount, sign, or both. For example, suppose that the figures composing the slowgrowing components are biased in an upward direction, while the figures representing the fastgrowing components are biased in a downward direction. The rate of growth of the aggregate
composed of these elements will be biased in a downward direction. All that can be done in this
case is to attempt to judge and to describe the direction and probable importanceof the bias in the
rate of growth.
The details of the construction of the capital stock estimates for the years 1774-1815 will be
provided in a monograph currently under way.
86
Robert E. Gallman
tion costs. The two are virtually identical, where it has been possible to run a
test. They are net of retirements and of capital consumption, with one exception, to be discussed further below.
The cost-of-living deflator is the one assembled by Paul David and Peter
Solar (1977), the only series that covers the full period. According to Claudia
Goldin and Robert Margo (1989), the index rises too little or falls too much
in the nineteenth century before the mid-1840s. If they are correct, the rate of
growth of the capital stock deflated by this series is too high in the period
before the mid-l840s, a point to which we will return. Dorothy Brady’s investment goods price indexes from volume 30 of Studies in Income and
Wealth (1966), extended to the years before 1840 in a variety of ways, were
the chief bases for the deflation of the capital stock, viewed as an input. The
Brady index numbers refer to census years. They had to be adjusted modestly
to make them relevant to the dates to which the capital stock estimates refer
(the last day of the census year). Conceptually, these index numbers are exactly what are required. They were augmented in various ways to permit the
deflation of inventories and certain types of farm improvements, for which
Brady supplied no indexes. The problems of assembling appropriate deflators
for the years before 1840 require a paper of their own. They are treated further
in appendix A.
2.3 Rates of Growth in Historical Perspective
The concern of this paper is with American economic growth before the
Civil War, which means that the measures of central concern to it are real
measures, particularly real measures deflated by population. The current price
estimates are worth at least a brief inspection, however. On the whole, they
are less processed than the real figures and may therefore be a little more
reliable. Table 2.2 contains current price estimates of the capital stock, conventionally defined.s Three points come through very clearly. The rates of
growth are all very high; the capital stock in 1980 was apparently about
40,000 times as large as the stock of 1774, an extraordinary figure. Although
most of the rates were computed over considerable stretches of time, and
therefore should not be unduly influenced by transient phenomena, they vary
quite widely from one period to another. Finally, it is clear that the experience
before the Civil War was by no means uniform. In particular, the rates of
growth are especially low in the years between the turn of the century and
1840, and especially high from 1840 to 1860. The second period is short, and
5 . The conventional concepts are the domestic capital stock and the national capital stock. The
former includes the value of structures, equipment, and inventoriesphysically located in the country at issue; the latter includes all of these items, but also adjusts for net international claims, so
that the measure includes the value of capital owned by nationals of the country at issue. Unconventional estimates may include, additionally, the value of the clearing and first breaking of land,
the value of human capital, the value of consumer durahles, etc.
87
American Economic Growth before the Civil War
Table 2.2
indexes and Average Annual Rates of Change of the U.S. Capital
Stock, Current Prices, 1774-1980
Domestic
Capital
Domestic Capital &
Consumer Durables
National
Capital
Panel A: Indexes
~
1774
1799
1805
1815
1840
1850
1860
1870
1880
1890
1900
1929
1953
1980
100
399
58 1
999
1,573
2,579
5,298
8,620
11,795
20,526
27,386
138,592
444,239
3,761,382
100
1,503
2,538
5,274
8,751
11,761
20,198
26,457
135,343
436,493
3,665,337
100
415
628
1,110
1,691
2,919
6,@33
9,201
12,805
22,396
30,886
170,360
541,061
4,560,608
Panel B: Average Annual Rates of Change (8)
1774-1840
1774-99
1799-1840
1840-1900
1840-60
1860-1900
1900-1929
1929-53
1953-80
1774-1980
4.3
5.7
3.4
4.9
6.3
4.2
5.8
5 .O
8.2
5.2
4.2
4.9
6.5
4.1
5.8
5 .O
8.2
5.2
4.4
5.9
3.5
5 .O
6.5
4.2
6.1
4.9
8.2
5.3
Sources: See text.
the rates of growth computed across it could be influenced by business cycles
or long swings. But Abramovitz’s (1989) chronology of long swings and protracted depressions suggests that this is probably not a problem.
The record described by table 2.2 is influenced both by real phenomena and
price level changes. The price index numbers in table 2.3 allow one to judge
how important the latter developments were. Between 1774 and 1900 the
long-term trend of the two price indexes appears to be close to zero, but in the
short term prices were quite unstable. In the twentieth century there is additionally a pronounced upward trend. Notice, finally, that while the two indexes tend to move together, the consumer index is the more volatile. The
plan to deflate by two separate price indexes, then, seems to have substantive,
as well as theoretical, merit.
88
Robert E. Gallman
Table 2.3
Capital Stock Deflators, Base 1860,1774-1900
Domestic Capital
Price Index
1774
1799
1805
1815
1840
1850
1860
1870
1880
1890
1900
1929
1953
1974
1980
81
111
115
157
91
94
100
127
112
96
90
165
357
Consumer
Price Index
97
148
141
185
104
94
100
157
123
109
101
205
320
589
1,193
Sources: See text.
The deflated series appear in table 2.4. Four matters of interest strike one
immediately: First, deflation does reduce the volatility of the series somewhat;
part of the short-term movement observed in table 2.2 is due to price fluctuations. Second, it is clear that the real capital stock has grown more slowly in
the present century than it had previously. Third, it is also clear that the rate
of growth accelerated between the years before 1840 and the years thereafter.
The broad pattern, then, is of an early acceleration, followed by a subsequent
retardation. Finally, notice that these findings emerge from all four series, the
national and domestic capital stocks, deflated by the consumer price index and
by the capital price index. But the detailed pattern of change differs from one
series to the other. For example, compare the results obtained for the period
1929-53. The real capital stock, viewed as accumulated consuming power,
grew much faster than did the real capital stock, viewed as an input to production: the prices of capital goods increased faster than consumer prices, between these two dates.
More interesting for present purposes is the pattern across the years 17741840. Notice (table 2.3) that consumer prices advanced much farther than
capital goods prices between 1774 and the turn of the century, and fell much
farther between then and 1840. Across the full span, 1774-1840, the two
index numbers show roughly similar changes, so that the two capital stock
series yield about the same results. But the interpretation of the subperiods
before 1840 depends entirely on the system of deflation one chooses to use.
And the systems of deflation, recall, view the capital stock in two quite different ways: as the value of the accumulations of the years, expressed in consumer goods, as against the productive power of the capital stock.
89
American Economic Growth before the Civil War
Table 2.4
Indexes and Average Annual Rates of Change of the U.S. Capital
Stock, 1860 Prices, 1774-1980
Domestic Capital Deflator
National Capital Deflator
Capital
Price Index
Consumer
Price Index
Capital
Price Index
Consumer
Price Index
100
289
409
513
1,401
2,212
4,292
5,486
8,462
17,217
24,552
68,472
102,132
223,632
100
262
400
525
1,472
2,665
5,148
5,335
9,318
18,295
26,347
66,398
137,182
100
306
449
57 1
1,514
2,497
4,849
5,897
9,157
18,665
27,632
77,681
114,109
297,638
100
27 1
43 1
58 1
1,571
3,007
5,805
5,669
10,071
19,877
29,584
80,390
163,571
4.2
4.6
4.0
5.0
6.0
4.4
3.0
3.6
1.6
3.6
4.0
4.3
4.1
4.4
5.0
6.8
4.2
Panel A: Indexes
1774
1799
1805
1815
1840
1850
1860
1870
1880
1890
1900
1929
1953
1980
Panel B: Average Annual Rates of Change (%)
1774-1 840
1774-99
1799-1 840
1840-1900
1840-60
1860-1900
1900-1980
1900-1929
1929-53
1953-80
1774-1980
4.1
4.3
3.9
4.9
5.8
4.5
2.8
3.6
1.7
3.5
3.9
4.2
3.9
4.3
4.9
6.5
4.2
3.2
3.1
3.5
3.O
Sources: See text.
No doubt the contrast is in some measure spurious, however. Items of construction compose an important part of the capital stock throughout (see table
2.8). The deflators for this component in the years before 1840 were constructed in part from data on wage rates. Wage rates tend to be less volatile
than prices (see Robert A. Margo’s paper in this volume). The capital stock
price index numbers for the period before 1840 may therefore understate the
fluctuations experienced by the prices of capital goods. It is thus possible that
the measured rate of growth of the real capital stock, viewed as an input, is
too high across the years 1774-99 and too low between 1799 and 1840. The
matter is unlikely to be important with respect to the main point of present
90
Robert E. Gallman
concern, however. It seems clear that the rate of growth of the capital stock
did accelerate between 1774-1840 and the subsequent years.
The capital stock treated so far ignores a component of investment that was
important, particularly in the years before 1840: the activities of land clearing
and first breaking which engaged so large a part of the working lives of American farmers (Primack 1962). Table 2.5 contains index numbers describing
the change over time in the real value of the domestic capital stock, inclusive
of the value of these farm-making activities. The overall rate of growth of this
aggregate-3.9 percent, 1774-1900-is very much lower than the one recorded for the less comprehensive capital stock treated in table 2.4-4.5 percent (capital stock deflator in each case). These findings reflect the fact, of
course, that farm formation was a very important part of capital, but one that
increased over time much more slowly than the other components of the stock,
a point to which we will return.
The acceleration picked out by the data of table 2.4 reappears in table 2.5
and in a more marked form. But notice that the pattern is somewhat different.
The series deflated by the prices of capital now shows a higher rate of growth
across the period 1799-1840 than across the period 1774-1799, in contrast to
the results shown by table 2.4. The explanation is that introduction of the
farm-making elements of the capital stock necessarily altered the capital price
index numbers. Farm making was carried out by farm laborers, and the value
of farm making is the value of the time of farm workers. Farm wage rates thus
figure in the estimation of the value of land clearing and breaking, as well as
in the deflation of these components of the stock. Farm wage rates rose quite
pronouncedly between 1774 and 1840, which gives the deflator an upward
tilt.
All of the series discussed above refer to the aggregate capital stock. A
more interesting variable, however, is the per capita capital stock. Estimates
appear in table 2.6. Deflating by population produces two important, if easily
anticipated, results. First, the retardation of growth in the twentieth century
disappears, while the acceleration between 1774-1 840 and 1840-1900 becomes very much more pronounced. The acceleration appears in every variant
but is particularly evident in the series describing the most comprehensive
measure, deflated by capital stock prices.
The acceleration in the rate of growth of the capital stock reflects in part the
increase in the investment rate and the rise in the capital/output ratio, which
seems to have begun as early as the turn of the century, at least in the case of
the conventional measurements, but which was particularly pronounced from
1840 until 1900 (Davis and Gallman 1978; table 2.7). That does not appear to
be the only source, however. The rates of growth of real national product per
capita from 1840 onward were higher than the rates of growth of real capital
per capita in the period before 1840, regardless of the capital concept adopted
and the deflator employed (Davis and Gallman 1978; Gallman 1966). Accepting the rate of change of the capital stock series before 1840 as an upper-
91
American Economic Growth before the Civil War
Table 2.5
Indexes and Average Annual Rates of Change of the U.S. Domestic
Capital Stock,Including the Value of Clearing and Breaking
Farmland, 1860 Prices, 1774-1900
Deflator
Capital
Price Index
Consumer
Price Index
100
227
290
353
913
1,362
2,432
3,004
4,520
8,491
11,807
100
245
332
379
1,229
2,140
3,980
3,884
6,543
12,229
17,253
Panel A: Indexes
1774
1799
1805
1815
1840
1850
1860
1870
1880
1890
1900
Panel B: Average Annual Rates of Change (8)
1774-1840
1774-99
1799-1 840
1840-1900
1840-60
1860-1900
3.4
3.3
3.5
4.4
5.0
4.0
3.9
3.1
4.0
4.5
6.1
3.7
Sources: See text.
bound estimate of the rate of change of real national product, the evidence
suggests quite clearly that the rate of growth of real national product per capita
accelerated in the years before the Civil War.
These results are generally consistent with Thomas Weiss’s inferences concerning income, which he derived from his labor force series (see table 2.7
and Weiss’s paper in this volume). Both Weiss’s figures and the capital stock
data were assembled from fragmentary evidence and are subject to substantial
margins for error. But both series seem to tell about the same story, and that
affords greater confidence that the story is a true one.6
6. The capital and income (Weiss) data permit a check on an inference advanced by Davis and
Gallman, who guessed that the net investment rate averaged between 6.2% and 7.0% in the period
1805-40 (Davis and Gallman 1978,2). The rates of growth and capital/output ratios in or underlying table 2.7 are consistent with net investment rates, computed against GDP, of between 5
percent and 6.5 percent. The Davis and Gallman figures were computed against NNP, however. If
the data in and underlying table 2.7 are adjusted to make them conform more nearly to the concepts Davis and Gallman were employing, the implied investment rates become roughly 5.9 percent and 7.2 percent, reasonably close to the Davis-Gallmanfigures.
92
Robert E. Gallman
Table 2.6
Indexes and Average Annual Rates of Change of the U.S. Domestic
Capital Stock and Structures, Per Capita, Conventional and
Unconventional Concepts, 1860 Prices, 1774-1980
Including Clearing & Breaking,
Deflated by
Conventional Concept,
Deflated by
Capital
Price Index
Consumer
Price Index
Capital
Price Index
Consumer
Price Index
Panel A: Indexes
1774
1799
1805
1815
1840
1850
1860
1870
1880
1890
1900
1929
1953
1980
100
132
154
143
193
224
32 I
323
396
643
759
1,348
1,520
2,735
100
120
150
147
202
210
384
315
436
683
815
1,461
2,294
100
112
125
106
169
217
100
104
109
99
126
138
182
177
212
317
365
229
306
456
534
0.4
0.1
0.5
1.8
1.9
1.8
0.8
0.4
1.o
1.9
2.9
1.5
291
Panel B: Average Annual Rates of Change (%)
1774- 1840
1774-99
1799- I 840
1840- 1900
1840-60
1860-1900
1900-1929
1929-53
1953-80
1774-1 900
1900-1980
1774-1 980
1.O
1.1
0.9
2.3
2.6
2.2
2.0
0.5
2.2
1.6
1.6
1.6
1.1
0.7
1.3
2.3
3.3
1.9
1.6
1.9
1.7
Sources: See text
2.4 Changing Composition of Capital Stock
Rates of change say something about the process of growth and development; data on the structure of the economy tell more. Development consists
of structural change.
The conventional measure of domestic capital, in current prices, exhibits
two pronounced compositional shifts: the fraction of the capital stock accounted for by animals drops very far, indeed, while the share attributable to
93
American Economic Growth before the Civil War
Table 2.7
Real GDP and Real Domestic Capital per Capita, Conventional and
Unconventional Concepts, 1840 Prices, 1800-1860
Real GDP per capita ($)
Conventional, variant A
Conventional, variant B
Unconventional, variant C
Real domestic capital per capita ($)
Conventional
Unconventional
CapitaVoutput ratios
Conventional, variant A
Conventional, variant B
Unconventional, variant C
1800
1840
-
73
66
78
91
91
101
125
125
135
104
175
157
219
262
316
1.42
1.57
2.24
1.73
1.73
2.16
2.09
2.09
2.34
1860
Sources: The real GDP per capita estimates are from Weiss’s paper in this volume. For the
remaining estimates, see the text.
structures rises, both of these developments occurring chiefly after 1815 (see
table 2.8). But current price data are not so useful, in this context, as constant
price data, which tell a very interesting story. They show that the structure of
the capital stock changed very little, down to 1840. Thereafter, there were
accelerating shifts. The share of animals in the total dropped precipitately and
inventories dropped mildly, while the share of structures rose a little and the
share of equipment rose very much. There is the strong suggestion of an economy shifting in the direction of industrial activity and modern economic
growth: away from agriculture and animal power, and toward manufacturing
and mechanical power. There is no question that stirrings can be identified
well before 1840-Kenneth Sokoloffs work shows clearly that important industrial change can be dated to 1820, at least. (See Sokoloffs paper in this
volume.) But these activities could not have carried a very heavy weight in
the economy much before 1840, and that is probably what the data in table
2.8 are showing us. Bias in the estimates may overstate the decline in the
relative importance of animals after 1870, and may contribute to the finding
of stability in the share of structures in the capital stock before 1840 (see
appendix A), but these matters are probably not of much importance.
The introduction of the value of farm making into the capital stock produces
some expected shifts. Concentrating on the constant price data, the value of
land clearing and breaking accounted for over half of the capital stock in 1774
and something under half in 1799. This figure dropped modestly to 1840when it was a little less than a third-and more dramatically thereafter, reflecting the relative decline of the agricultural sector. In this variant, inventories retained roughly the same share of the capital stock after 1799, while the
share of structures experienced a strong upward movement from the same
date.
94
Robert E. Gallman
Constituents of the Domestic Capital Stock, Expressed as Shares in the
Domestic Capital Stock, 1774-1900
lsble 2.8
1774
~~~
1799
1805
1815
1840
1850
1860
1870
1880
1890
1900
.54
.I2
.22
.I2
.54
.I1
.24
.I1
.55
.lI
.24
.09
.61
.I3
.I9
.08
.60
.I4
.I9
.07
.54
.I2
.22
.I2
.55
.I3
.22
.I0
.50
.I6
.25
.09
.49
.25
.21
.06
.46
.30
.I9
.04
.42
.09
.I7
.09
.22
.44
.09
.20
.09
.I7
.47
.I0
.21
.08
.I4
.55
.I1
.55
.I3
.I7
.07
.I0
.I8
.42
.44
.09
.I7
.09
.I1
.I8
.41
.I3
.21
.07
.I7
.44
.22
.I9
.05
.42
.28
.I8
.04
.08
~
Panel A: Excluding the Value of Farmland Clearing and Breaking
Structures
Equipment
Inventories’
Animals
.39
.I3
.23
.25
.33
.I4
.35
.I8
Structures
Equipment
Inventories’
AnimaIs
.40
.08
.28
.25
.34
.09
.35
.23
.35
.I5
.34
.I6
Current Prices
.41
.45
.47
.I3
.I4
.I3
.26
.24
.26
.21
.I7
.I3
Constant (1860)Prices
.40
.09
.32
.I9
.41
.07
.29
.22
.43
.08
.26
.23
.46
.09
.27
.I7
Panel B: Including the Value of Farmland Clearing and Breaking
Structures
Equipment
Inventories’
Animals
Land clearing &
breaking
.24
.08
.I4
.I5
.40
.21
.09
.23
.I1
.36
.26
.I1
.24
.I2
.28
Structures
Equipment
Inventoriesa
Animals
Land clearing &
breaking
.I7
.I9
.04
.05
.I2
.I1
.56
.20
.I3
.44
.25
.05
.20
.I2
.39
Current Prices
.33
.33
.35
.10
.I0
.I0
.21
.I8
.20
.I7
.I2
.I0
.19
.28
.25
.06
.08
Constant ( I 860) Prices
.27
.05
.I9
.14
.36
.29
.06
.I7
.I5
.32
.33
.07
.I9
.I2
.28
.22
.08
.I9
.I1
Sources: See text.
aExcludinganimals.
Table 2.9 is another way of considering the same phenomena. It shows
indexes of the per capita supply of each of the components of the capital
stock. The growing importance of structures and, particularly, equipment
comes through powerfully, while the value of the stock of land clearing and
first breaking is shown to have fallen well behind the growth of population.
There were two elements involved in the production of this result. First, the
volume of farmland per capita declined over time, as the population became
less and less rural and farm-centered. Since American agriculture was able to
feed a growing population and expand its overseas sales, the decline in the
value of farm improvements per capita went hand in hand with the growing
95
American Economic Growth before the Civil War
Indexes of Per Capita Real Magnitudes, 1860 Prices, 1774-1900
Table 2.9
Structures
Equipment
Inventories’
Animals
hdclearing &
breaking
1774
1799
1805
1815
1840
1850
1860
1870
1880
1890
1900
100
100
100
100
112
142
166
122
81
156
166
176
121
74
150
133
211
202
178
179
73
263
262
218
154
70
438
479
253
154
72
449
538
258
126
62
503
785
360
139
66
793
1,981
479
148
60
886
2,867
526
132
55
100
149
130
64
Sources: See text.
‘Excluding animals.
productivity of agricultural land. Second, as population moved westward, out
of the wooded areas, the cost of preparing land for cultivation fell. Toward the
end of the nineteenth century, then, the real value of farm improvements (exclusive of structures) per acre was smaller than it had been in the eighteenth
century. The meaning of this change is taken up further in appendix A.
On the whole, the structural evidence supports the conclusions that one
might tentatively draw from the aggregate series: the American economy began to experience the process of modern economic growth in the years after
the War of 1812; by the 1840s the modem components of the economy were
large enough and growing rapidly enough to have an observable impact on the
rate of growth and the structure of the economy.
2.5 The Growth of Total Factor Inputs
The measurements of the capital stock, viewed as an input to the productive
process, yield information that clearly bears on the speed and nature of American economic growth. Measurements of total factor inputs would be even
more useful. The assembly of the additional required inputs is not very difficult. Estimates of the volume of agricultural land (the only land input that
could be taken into account) already exist. (Gallman 1972, 201, 202, extended to 1774 in the same manner as the extension to 1800.) Weiss has generated new labor force figures for the years 1800-1900, at ten-year intervals,
and they were readily extended to 1774.’
7. The estimate is based on Jones (1980, 30) and Weiss’s chapter in this volume. According to
Jones there were 53,056 indentured servants in 1774 and 480,932 slaves. All indentured servants
were in the work force; following Weiss’s judgment for 1800, slaves ten and older probably
amounted to 65 percent of the population of slaves, and nine-tenths of these people were in the
work force. According to Jones, there were 396,158 free adult males, of whom, if we follow
Weiss’s treatment for the nineteenth century, 87.2 percent were in the work force. The rest of the
population-I ,034,456-consisted of youths and children, by Jones’s account. Assuming half
were males (a safe guess) and that they were distributed among the age groups as the white population of 1800 was, then there were about 55,000 males, ten to fourteen years old, of whom 22.1
96
Robert E. Gallman
Table 2.10, panel A, contains statements of the rates of growth of each
input and each input per member of the population for various periods between 1774 and 1900. Notice that the labor force grew slightly more slowly
than population between 1774 and 1800 and a little faster between 1800 and
1840. Thereafter, with the expansion of immigration, and its effect on the
structure of population, the labor force participation rate rose faster than before. On the whole, the patterns of change of the other inputs are similar. The
volume of agricultural land per capita actually declined throughout, but the
rate of decline was less after 1800 than before, while the quantity of capital
increased faster than population, the rate rising persistently over time. The
strong suggestion of these data is that the per capita supply of all inputs, taken
together, must have grown very slowly, if at all, down to 1800, when it began
to increase, the increase becoming more marked as time passed.
This, in fact, is what is shown by panel B of the table, which sets out the
rates of change of all three factors combined. The rates of growth of total
inputs and inputs per capita accelerated over time, the change in the per capita
rates being particularly striking.
There are three series describing rates of change of aggregate inputs. In the
first, the underlying labor input is measured by the numbers of workers, without regard to the length of the work year or the differential quality of the
workers. In the second and third, very crude efforts have been made to adjust
the labor supply for sectoral differences in the work year, trends over time in
the work year, and differences among sectors in the “quality” of workers. In
series LFQV, the weights by which the rates of change of the three input series
are combined (estimated factor income shares) vary from one year to the next;
in series LFQF, the weights are fixed at the 1880 levels. The techniques employed to make the estimates are described in appendix B; the adjustments are
almost certainly too large. That is, the rates of change represented by
LFQV-and possibly LFQF, as well-are probably too large. The three sets
of figures, however, may very well establish boundaries within which the rates
of change of a properly adjusted labor input series would lie.
In any case, the rates of change of the combined input series do describe
the same general pattern: an acceleration in the supply of inputs and, especially, inputs per capita. For the period following 1800, these findings once
again parallel Weiss’s (table 2.7). Furthermore, there was not only an acceleration in the rate of change of aggregate inputs, but also in total factor productivity: the long-term rate of gain was substantially higher after 1840 than
before (table 2.10, panel C).
percent were in the work force (following Weiss’s judgment for 1800), and there were 53,815 who
were fifteen to twenty years old, of whom (again following Weiss) 87.2 percent worked. Adding
free females, ten years old and older (497,973, with a participation rate of 7.5 percent, per Weiss),
brings the total labor force to 776,241. A check on the total: assuming an overall participation rate
of 32.5 percent (typical of the early decades of the nineteenth century, according to Weiss) yields
a figure of 765,039, close enough.
97
American Economic Growth before the Civil War
Table 2.10
Rates of Growth of Factor Supplies, Factor Supplies per Capita, and Total
Factor Productivity, 1774-1900
~
1774-1800
1800-1840
1840-1900
~~
1840-60
Panel A
Labor force (LF)
LFIpopulation
Land
Land/population
Capital (K)
Upopulation
3.09
-0.08
2.26
-0.91
3.39
0.22
3.09
0.11
2.80
-0.18
3.45
0.48
2.72
0.20
2.17
-0.35
4.40
1.88
3.41
0.31
2.87
-0.23
5.17
2.07
3.10
-0.07
3.21
0.04
3.25
0.08
3.18
0.20
3.44
0.46
3.47
0.49
3.20
0.68
3.75
1.23
3.57
1.05
3.91
0.81
4.78
1.69
4.41
1.31
.80
.25
.43
.82
.70
-.17
.20
Panel B
Total factor inputs, LF
Total factor inputs/population, LF
Total factors, inputs, LFQV
Total factor inputs/population, LFQV
Total factor inputs, LFQF
Total factor inputs/population, LFQF
Panel C
Total Factor Productivity
GDP, LF
GNP, LF
GDP, LFQV
GDP, LFQF
.46
Sources: The real GDP estimates underlying the first set of total factor productivity estimates (panel C)
are Weiss’s, chapter 1 in this volume (broad concept, variant C). They are expressed in 1840 prices, as
are the capital stock estimates (domestic capital) used with them to estimate total factor productivity.
The real GNP estimates (panel C) were derived from those underlying Gallman (1966). They are
expressed in 1860 prices and include the value of all land improvements made in the given year and the
value of home manufactures. The capital stock estimates used in the analysis involving the GNP refer to
the national capital stock.
The labor input series is based on Weiss’s labor force figures. LF refers to this series in unadjusted
form. LFQV means that the labor force has been adjusted to take into account differences in work time
and labor quality, both among sectors and over time (1840 onward); that is, the sectoral “weights” are
variable. LFQF means that the labor force figures have been adjusted to take into account differences in
time and quality among sectors, but not across time; that is, the sectoral “weights” are fixed. (In fact,
the weights employed are those of 1880; only two sectors are distinguished in the fixed weight variant:
“agriculture” and “all other.”)
The rates of growth of the capital stock, 1840-1900, were computed from the series that incorporates
the value of fencing.
The weights assigned to the rates of growth of the individual factors of production are labor, .68; land,
.03; and capital, .29. These weights are intended to reflect income shares. (Land improvements, of
course, are treated as capital.)
For estimating details, see the text, especially appendix B.
98
Robert E. Gallman
These results are surely not amazing. The years from 1774 through 1815
were years in which the young country engaged twice in major wars; when
peace was achieved, American products were frequently prevented from entering their natural markets under reasonably free conditions. There was one
period of booming trade, when the Napoleonic Wars created great opportunities for American merchants, opportunities ended by the Embargo of 1807
and then the War of 1812. With the return of peace, the factory system began
to spread in earnest, and by 1840 the production of textiles had been virtually
completely transferred out of the home and the shop and into the factory. The
variety of American manufacturing activities increased markedly in the 1840s
and 1850s, and machine building began to assume the central position it was
to occupy in American industrializationfor the rest of the century. The aggregate statistics are simply the embodiment of these well-known developments.
The degree to which the benefits of economic growth were offset by costs
unrecorded here and the extent to which the benefits were shared among
Americans are matters of considerable importance. But since they are taken
up by John Wallis and me in the Introduction and by other authors contributing
to this volume, it is reasonable to pass them by here.
2.6
Conclusions
The conclusions of this paper are readily summarized. The capital stock
series suggest that the pace of American economic growth accelerated in the
decades before the Civil War. The evidence for this statement is to be found in
the real per capita capital stock figures, the various estimates of aggregate real
inputs per capita, and the changing structure of the capital stock, which describe a process of industrialization. The components that make up the series
have their weaknesses, but the review of these components conducted above,
and also in appendix A, turned up no compelling reasons to believe that the
computed rates of growth and structural changes are importantly biased.
The acceleration of the rate of growth should not be allowed to obscure the
progress made before 1840. The series assembled in this paper support
Thomas Weiss’s finding that per capita GDP increased in the decades between
1800 and 1840. Furthermore, the per capita supply of capital seems to have
been increasing since 1774, and the supply of all factors of production, combined, seems to have increased at least as fast as population between the beginning of the Revolution and the turn of the century. There were bad times as
well as good ones, and the standard of life surely sometimes declined, perhaps
for extended periods. But the trend was mildly favorable between 1774 and
179911800, if these series are to be believed, more clearly favorable from the
turn of the century until 1840, and even more pronouncedly favorable thereafter. The capital stock figures, however, bear only on the side of life that has
to do with the provision of commodities and services. Industrialization may
99
American Economic Growth before the Civil War
have brought a deterioration of the quality of life for some and may have for a
time overwhelmed the capacity of society to deal with problems of public
health. Other indexes of the standard of life, stressing health, for example,
may yield results at odds with those reported in this paper-certainly this is
the suggestion of the work described by Richard H. Steckel in his essay for
this volume. The important point to be taken from the results described
herein, however, is that the performance of the economy, narrowly conceived,
was improving, and at an accelerating pace. The means for dealing with the
problems created by the reorganization of society were therefore increasing.
Solutions awaited the accumulation of the necessary knowledge and the emergence of a will to act.
Appendix A
Estimating Problems and Tests of Estimates
This appendix takes up a few of the chief problems encountered during the
construction of the capital stock estimates, and describes some of the tests
that were run to check the estimating decisions that were made.
Land Clearing and Breaking
The largest item in the more unconventional-but more meaningful-of
the capital concepts employed in this paper is the value imparted to land by
the processes of clearing and first breaking. The estimating procedure was
simple. The following variables were established for each year: the number of
acres of improved farmland of each relevant type (land originally under forest,
land originally under grass) in each state or region; the number of labor hours
per acre required to improve land of each type; the cost of farm labor in each
state or region (Primack 1962; Lebergott 1964). Simple multiplication and
addition produced the final figures. Constant price estimates were obtained by
substituting technical coefficients and wage rates relevant to 1860 for those
relevant to the current year. For the years 1840-1900-but not earlier-estimates of the value of fencing, drainage, and irrigation works were also made.
Certain characteristics of the series that may be associated with biased rates
of change are immediately evident. The weight attributed to the clearing and
breaking series is incorrect; it is probably too low, especially for the years
before 1840. Since the clearing and breaking series exhibits relatively low
rates of change over time, giving it a heavier weight would tend to reduce the
rates of growth of the aggregate capital stock series, particularly before 1840.
Thus the acceleration of the rate of change described previously in this paper
would be enhanced.
100
Robert E. Gallman
The weight attached to the series is too low because the estimates ignore all
elements of clearing and breaking cost except labor. Labor was, no doubt, the
principal cost, but it was not the only one. Second, the only improved land
treated is agricultural land; no account is taken of land under houses, factories, shops, and so forth. Third, for the years before 1840, important elements
of improvement-particularly fencing-had to be ignored. If it had been possible to treat all of these phenomena, the improvement series would have had
a larger weight.
There are, however, certain offsets. First, the value of fencing may very
well have increased faster than the value of clearing, before 1840; it is almost
certainly true that the volume of land under houses and so forth increased
faster than the volume of improved land in agriculture, at least after 1840.
Introducing these elements into the analysis might raise the rate of change of
the improvements series, although probably not by much.
Another factor may appear, at first blush, to be more important than any so
far discussed: the estimates make allowance for land retirements (land allowed
to go back to nature), but not depreciation. The reason depreciation has been
ignored is that land improvements, if properly maintained, do not depreciate.
Bad farming practices may erode the fertility of the land, and the opening of
western farms may reduce the value of eastern farms, but these changes have
to do mainly with the value of land, rather than with the value of improvements. Now in a sense this characteristic of improvements is shared with other
elements of the capital stock. Properly maintained, houses and ships and even
machines can last very long, indeed. The difference is that most of the houses,
ships, and machines that existed in, for example, eastern Pennsylvania in
1774 are gone today, while much of the improved land of that period is still
improved. A substantial part of it is now under houses and shopping malls and
highways, rather than under Indian corn, but it is still improved. Furthermore,
in the cases of buildings, machines, and so forth, one can devise reasonable
depreciation rates that properly describe the average lifetime experiences of
these elements of capital and that are roughly relevant to long reaches of history. That is not possible for land improvements.
The discussion above implicitly introduces another issue. The improvements series consists of reproduction cost estimates. Various tests have shown
(see Gallman 1987) that the reproduction cost and the market value of structures and manufactured producers’ durables were, on average, about the same
in the nineteenth century. Is this also true of land improvements? If not, then
how is the analysis affected? The few simple tests that have been run seem to
suggest that they are alike. At least two efforts have been made to estimate the
market value of clearing and breaking at midcentury: one by Stanley Lebergott for the Midwest, the other by Stanley Engerman and Robert Fogel for the
South (Lebergott 1985; Fogel and Engerman 1977). Comparisons are not easily made, and the efforts reported here may be polluted by wishful thinking,
101
American Economic Growth before the Civil War
but I do not believe that is the case. The results suggest that estimates computed along the lines laid out above are very similar to the ones obtained by
Lebergott and Engerman and Fogel. The suggestion then is that the market
price and the reproduction cost of land improvements were about the same,
on average, at midcentury.
The same may also hold for 1774. At least it is true that when one subtracts
from Alice Jones’s estimate of the value of real estate, my estimates of the
value of land clearing and structures plus a rough allowance for other elements
of land improvement (a relatively small part of the total), the remaining value,
divided by the number of acres of land privately held (derived from Blodget
[ 18061 1964), yields an average price of land per acre-exclusive of improvements-that is almost identical with Blodget’s estimate of the average value
of unimproved land in 1774. The test is very roundabout and places much
weight on a residual. Nonetheless, it encourages one to think that market price
and reproduction cost may have been about the same, on average, at that
date.
There is some evidence to the contrary, however. Specifically, Blodget’s
estimates of the average value of improved land per acre in 1774, 1799, and
1805 are substantially smaller than my estimates of the cost of improving
land per acre. Bear in mind that Blodget’s figures include the value of the land
itself, while mine do not. The margin is so great that one has the impression
that if Blodget’s figures are truly market-price figures, and mine truly reproduction cost figures, then farmers of the late eighteenth century and the early
nineteenth century were behaving irrationally, improving much more land
than could be justified by the market. I do not believe that and therefore think
that either Blodget is wrong or I am.
In making my estimates I assumed that all of the land improved at each of
these dates had originally been forest land. That is probably not correct, and
since forest land cost more to improve than grass land, this assumption probably leads to an overstatement of the value of cleared land at these dates. But
the overstatement is tiny and is surely more than offset by the fact that the cost
of factors other than labor was left out of account.
I also assumed that the labor hours per acre required for clearing were the
same at these early dates as at midcentury. Primack (1962) believed that there
were no important improvements in clearing techniques until after the Civil
War, and while his interests were confined to the last half of the century, his
remark is probably relevant to the early dates treated in this paper as well. In
any case, if I am wrong about this matter, I have understated the value of
improvements at these dates, not overstated them.
I also assumed that the treatment of stumps was the same at all dates: specifically, that one-third of the stumps were removed immediately and that the
rest were left in the land to rot away on their own. It may be that an even
smaller share of the stumps was taken out in the earlier years, but allowing for
102
Robert E. Gallman
the removal of no stumps would not bring my estimates and Blodget’s very
much closer together.8
A more promising source of disparity lies in the way in which labor time
was valued. I assumed that the opportunity cost of the labor employed in
clearing and first breaking could be approximated by the agricultural wage
rate. In fact, however, one would suppose that clearing and first breaking
would have been conducted by farmers in the off season, when real opportunities may have been restricted to maintenance tasks around the farm, hunting, fishing, and so forth. The wage rate, then, may overstate the opportunity
cost of labor. That seems not to have been the case at midcentury, when, as
indicated above, reproduction cost and market value of improvements were
very similar. It may be that by midcentury clearing and breaking were more
commonly hired out (e.g., to prairie sodbusters) than previously and that
farmers themselves had better opportunities for off-season work. If that were
the case, the estimating technique might work better for the mid-nineteenth
century than for the earlier dates. But that would be a relatively unimportant
matter. Our concerns are chiefly with the constant price series, which are
properly a function of the techniques and wage rates of 1860. The contrast
with Blodget refers exclusively to the current price estimates.
In any case, my by no means unbiased guess is that Blodget is simply
wrong on the matter of the value of improved land. The check of my work
against Jones’s estimate of the value of real estate and Blodget’s estimate of
the value of unimproved land seems reasonably strong. Furthermore, in comparison with Jones’s estimate, Blodget’s figures on the values of improved
land seem very much too low. I therefore incline toward the view that the
improvements series-particularly in constant prices-gives a reasonable
view of what it purports to describe. At least I cannot make a case for viewing
the series as strongly biased in one direction or the other or as generating
strongly biased rates of g r ~ w t h . ~
Structures
The estimates for 1850-1900 rest chiefly on census data; for 1840 on the
work of Seaman (1852); for 1815 on the direct tax of 18 13-1 5 and the work
8. The matter of stumps is tricky. What is the reproduction labor cost of ten acres of stumpless
cleared land that was formerly under trees? Is it the full labor cost of clearing the land and removing all the stumps? Or is it the labor cost of cutting down the trees, removing the one-third of the
stumps that were originally removed, and plowing the land? I decided that the second choice was
the correct one, but clearly one could make a case for the first, or perhaps even a third or fourth
option.
9. A word should be said about the land series, although there is inadequate space to go through
the estimating procedures and tests. The 1850-1900 data come from the census, with some adjustments. The adjustments depend in part on the work of Primack (1962). The 1840 figures are
weaker. They come from Seaman (1852) again adjusted and distributed, partly on the basis of the
work of Primack. The figures for 1774 through 1805 are from Blodget ([I8061 1964), adjusted in
various ways. The 1815 figure is a rough extrapolation from 1805. For a discussion of these
matters, see Gallman (1972).
103
American Economic Growth before the Civil War
of Pitkin (1835); for 1805 on the work of Blodget ([1806] 1964) and Goldsmith (1952); for 1799 on the direct tax of 1798 and the work of Soltow
(1984); and for 1774 on the work of Jones (1980). All of these data have been
very heavily processed, frequently with the object of extracting one element
from a larger aggregate, or dividing the aggregate among its components. In
each case but two, however, there is a quite substantial component of real data
that bears directly on the estimating problem. The weakest links are the ones
for 1805 and 1840; there are no data expressly relevant to these dates. The
underlying sources of evidence are the works of Seaman and Blodget. The
latter extrapolated his estimate from an earlier date, for which real evidence is
available, while the former both extrapolated from an earlier date and blew up
partial estimates to encompass the universe. These figures have been tested,
of course, but they are less trustworthy than are the rest.
There is not space to deal with all the estimating problems and with all the
tests run with respect to the estimates relating to structures. In what follows,
the most serious problem, which has to do with deflation for the years before
1850, will be treated.
For the years 1850-1900 there is no serious problem relating to deflation;
indeed, the price index number situation is unusually good. For most of these
years Dorothy Brady’s two sets of deflators-for houses and churches, on the
one hand, and factories and office buildings, on the other-are available.
These are true price indexes, which makes them quite unusual among construction deflators. Usually it is necessary to make do with cost indexes. Brady’s data need modest adjustment to make them expressly apposite to the task
of deflating the capital stock, but no heroic efforts are needed to put them in
proper condition for this purpose.
The problem appears in the years before 1850, for which Brady’s indexes
are not available. One possibility for this period is to follow the lead of David
and Solar (1977), who linked Brady’s housing price index to a construction
cost index and then carried it back to the late eighteenth century. Since the
relative importance of factories and office buildings before the 1840s was
probably slight and since construction techniques in this period may not have
varied much between residential construction and commercial buildings (except at the cutting edge of factory design and construction), an extension of
the housing price index would be an entirely adequate way to deal with the
deflation problem for all kinds of structures. David and Solar, however, did
not use Brady’s published series; they used the unrevised figures that Brady
prepared for the Income and Wealth Conference. It turns out that in most
instances the differences between the published and unpublished series are
slight-matters of a point or two. There is one exception. In the published
conference volume, Brady (1966) dropped her estimate of the price index of
housing in 1839.
The Brady unpublished index falls from a level of 128 in 1839, to 94 in
1849, and then rises to 100 in 1859. Available construction cost indexes fall
104
Robert E. Gallman
much more modestly and rise more sharply over these two decades, implying
that, if the unpublished Brady index is correct, productivity in construction
must have been rising quite dramatically. David and Solar believe that the
experience reflects chiefly the diffusion of the balloon frame, which was invented in 1833. They therefore suppose that the annual rate of productivity
improvement realized in the 1840s was also achieved in the period 1834-39.
They construct a building cost index and employ it with the Brady price index
to estimate productivity gains, 1839-59, and they then use it, together with
their estimate of the rate of productivity improvement, 1839-49, to extrapolate the Brady price index number for 1839 back to 1834. They assume that
there were no important productivity improvements before 1833 and extrapolate the 1834 price index number to earlier years in the century on their construction cost index. The productivity improvement for the period 1834
through 1859 implied by their calculations is a little more than 36 percent.
The procedure is ingenious and surely adequate to the purposes of David
and Solar. It is not so clear that it is adequate to the purpose of creating a
deflator for the most important component of the conventional capital stock
series. First there is the matter of Brady’s decision to suppress her 1839 estimate. Does this mean that she had had second thoughts about the strength of
that estimate? Presumably. Nonetheless there remains evidence that Brady believed that construction prices did fall in the late 1830s and early 1840s. Her
price index for factories drops very sharply between 1836 and 1844, for example. But of course this index refers to factories, not residences.’O
Is it reasonable to suppose that the balloon frame led to a rise in productivity of 36 percent in the first twenty-five years of its existence? Probably not.
The balloon frame saved on framing. Framing accounted for about 25 percent
of the cost of a building. Consequently, even if the balloon frame eliminated
the expense of framing and even if the balloon frame was adopted throughout
the industry within this period, the rise in productivity could not have come
close to reaching 36 percent. Neither condition was met, of course.”
10. One should not infer much about productivity changes from the relative movements of price
and cost indexes between 1836 and 1844, however. Between these two dates lay a very sharp
contraction. At least part of the decline in prices reflected falling profits, not rising productivity. It
is also likely that workers discounted standard wage rates in order to hold their jobs.
11. For example, “although many authorities assert that balloon frame construction had ‘almost
completely replaced the hewn frame for domestic construction by the time of the Civil War’ . . .
in North Carolina field surveys demonstrate the prevalence of heavy mortised-and-tenoned house
frames until the Civil War” (Bishir, Brown, Lounsbury, and Wood 1990,457). An architect whose
book was published in 1855 writes: “There is no doubt that if the subject received closer attention,
a better mode of framing than that generally employed, could be suggested. Timbers are often
unnecessarily heavy, but are afterwards so weakened by the mode of framing which is in vogue,
and which compels the cutting of mortices and tenons and insertion of one timber into another,
that the frame is less substantial than if constructed of lighter stuff differently put together. It is
difficult to persuade carpenters of this” (Wheeler 1855,407). The implication of the last statement
is important. The building industry was a conservative, locally organized industry. The architect
goes on: “The New York Tribune of January 18th. 1855, reported a meeting of the American
Institute Farmers’ Club, and contained amongst other items some remarks from one of the mem-
105
American Economic Growth before the Civil War
The framing of a building called for many workers. Barn-raising parties
were organized expressly for this purpose. The balloon frame eventually
changed all that. With the new system a man and a boy could frame a house
by themselves. Thus the innovation became immensely important to the farming community, particularly for people on the frontier, for reasons that transcended normal cost considerations. It also diffused quickly in new western
cities, places under intense demand pressure and without established artisanal
power groups. (Chicago and San Francisco were both balloon frame cities.)
But it did not immediately spread to the East.
There were, of course, other innovations during this period, so that the rise
in productivity that David and Solar identify need not be the result exclusively
of the balloon frame. The principal changes that seem to have been taking
place involved the transfer of some activities from the building site to mills.
For example, it is said that it became more common to use manufactured
nails, as well as manufactured windows and doors, which presumably lowered costs. But the census returns of 1810, 1850, and 1860 suggest that manufactured nails were already widely used before the 1830s. Mill-made sash,
doors, and blinds do not appear in the census returns-separately, at leastbefore 1860, when their output amounted to a value of about $9.5 million, in
a year in which the total value of conventional construction (exclusive of railroads and canals) ran to about $345 million. Mill-made windows and so forth
were therefore by no means negligible by this date, but they did not bulk large
enough to suggest that their introduction led to a major improvement in productivity. Furthermore, it may well be that their contribution to productivity
actually came after 1849, rather than before. At least the treatment of these
lines of production by the census suggests that this was so. David and Solar
find most of the productivity change (almost three-quarters of it) occurring
before 1849.
The general idea lying behind the David and Solar treatment of construction
prices is clearly reasonable, and their execution of it may have solved their
problem satisfactorily. The technique is less likely to solve my problem satisfactorily, however. Unfortunately, there is no option that is clearly superior.
bers upon a novel mode of constructing cheap wooden dwellings” (408). The “novel method” was
the balloon frame.
The extent to which innovations had diffused is relevant because it would have determined the
degree to which prices responded to innovations. Prices would have been potentially affected only
in localities in which the new framing system had begun to diffuse, and even there, prices need
not have fallen immediately, if competition among builders was not severe. If builders commonly
used cost plus pricing, of course, prices would have fallen immediately in areas where the balloon
frame was put in use.
There is a question as to whether Brady’s prices refer to average practice or best practice. I have
assumed they refer to average practice. If I am not correct in this assumption, and if builders
followed cost plus pricing practices, then the Brady price index numbers exaggerate the true
decline in average prices. The course of average relative prices of residences after 1849 suggests
that the ambiguity with respect to the meaning of the price indexes is unimportant for these years.
106
Robert E. Gallman
Nonetheless, I decided to accept the Brady indexes for the years 1849 onward.
I then adjusted them to fit my needs, and extrapolated the adjusted 1849
(1850) index number to 1840, 1815, 1805, 1799, and 1785 on the Adams
(1975) variant B (allowing for input substitutions) construction cost series.
The index was extended to 1774 on a construction cost series based on the
David and Solar common wage index, a Maryland farm wage rate, taken from
Adams (1986, 629-30), and the Bezanson-Gray-Hussey arithmetic average
price index for Philadelphia (U.S. Bureau of the Census 1975, ser. E-1 11).
The last two steps need further discussion.
The Adams construction cost index was exceptionally carefully made from
good basic data. It is an excellent construction cost index, and the version
used allows for factor substitutions due to shifts in relative input prices. For
present purposes, however, it has certain potential shortcomings. The ideal
index, for present purposes, is a true price index, an index that allows for
changes in productivity. The Adams index does not do that, except insofar as
productivity changes are associated with shifts in factor proportions. As proxy
for a true price index it will exaggerate any long-term price increases and
understate any long-term price decreases, so long as productivity improvements are taking place. The capital stock series that it is used to deflate will
then exhibit a rate of change that is biased in a downward direction. In the
present instance, the bias would exaggerate the observed acceleration in the
rate of growth of the capital stock. If the bias were serious enough, it would
account fully for the acceleration. That seems highly unlikely, however. The
sources of productivity improvement in construction do not appear to have
been important before the mid-l830s, and, as I have tried to show, even in the
period between the mid-1830s and the beginning of the true price indexes in
1849, the amount of productivity improvement is unlikely to have been very
great. In any case, the Adams index has other shortcomings for present purposes, and it turns out that at least one of these may introduce a compensating
bias, in direction at least, and perhaps in amount as well.
The Adams index refers exclusively to Philadelphia. How successfully does
it represent the United States? Two questions immediately arise. First, housing price levels varied by region, and as time passed, the relative importance
of the various regions changed. Did the shifts in regional weights affect the
trend in the national average of housing prices? Probably not, and if they did,
they tended to raise average prices a little. By ignoring the effects of the regional shift I can perhaps compensate slightly for whatever bias is, present
from the use of a cost index in place of a true price index. These conclusions
are based on the results of a test of the following form.
The census of 1840 requested information on the numbers of two types of
houses constructed in the census year, those built of brick and stone and those
built of wood, as well as the value of both types of houses taken together. I
used the state data in a regression analysis to obtain an intercept value and
coefficients for each of the two types of houses. These data were then em-
107
American Economic Growth before the Civil War
ployed to value the houses constructed in each state, and the figures thus obtained were divided through the census returns of the value of houses built to
get an index number for each state. The index number compares the value of
the houses constructed in the state with the value that would have obtained if
prices had been at the level of the national average. Clearly, the index numbers
reflect not only variations in building prices-which are required for the proposed analysis-but also differences in average size and quality of new
houses, from state to state. Since cost, size, and quality were likely to have
varied together-frontier areas having lower building costs, smaller houses,
and houses of lower quality than the well-settled areas-the index numbers
almost certainly exaggerate the regional variations in building costs, a point
to be borne in mind as the analysis unfolds.
The individual state index numbers were then used to deflate the state returns of the value of real estate in 1799, according to the direct tax. The sum
of the deflated returns was then divided through the aggregate current price
value of real estate in 1799, according to the direct tax. The result was an
index number of 0.932, which compares with the 1840 index number of
1 .OOO; that is, according to these calculations, the shifting weights among
states tended to raise, very slightly, the true price index of structures between
1799 and 1840.12 The index numbers almost certainly overstate the true impact of the redistribution of the value of structures among states in this period,
because the state index numbers probably overstate (for reasons previously
given) the true variation in building costs among states. It appears unnecessary, then, to adjust the Adams cost index to take into account the effects
of the shifting real-value-of-structuresweights among states. This is particularly the case in view of the fact that the Adams index is a cost index and is
likely, therefore, to exaggerate the extent to which the prices of buildings rose
or to understate the extent to which they fell during this period. Finally, if the
bias is slight between 1799 and 1840, it is almost certainly negligible between
1774 and 1799.
There is another aspect of the regional specificity of the Adams index that
must be considered. Do changes in Philadelphia costs properly represent
changes in costs in other regions? The strong suggestion that one gets from
looking at price and wage indexes from New England and New York (Rothenberg 1988; David and Solar 1977; Warren and Pearson 1933) is that they do
not: Adams’s cost index moves in step with the Bezanson-Gray-Hussey general price index (Philadelphia-U.S. Bureau of the Census 1975, ser. E-97),
while the Rothenberg, David-Solar, and Warren-Pearson indexes also move
more or less together, but quite differently from the Philadelphia indexes. (At
least these statements apply to the benchmark dates relevant here.) David and
12. The two indexes should ideally be weighted by the state distribution of the real value of
houses in the capital stock. These, in fact, are the weights utilized for 1799, but the weights for
1840 are the real values of houses built in the census year.
108
Robert E. Gallman
Solar report that a construction cost index they assembled from materials
prices from New York (Warren and Pearson 1933) and common wage rates
from Philadelphia (Adams 1975) and the Erie Canal (Smith 1963) exhibits a
less pronounced decline between 1809 and 1834 than does the Adams index.
I constructed an index from Warren and Pearson materials prices and DavidSolar common wages (using Adams’s weights and his procedure for allowing
for factor substitutions) for all the relevant dates. The Adams index shows a
much more pronounced drop over time than does the WP-DS index. There is
the strong suggestion that a properly derived national construction cost index
would exhibit more pronounced price increases and less pronounced price
declines, over the long run, than would a Philadelphia index. The bias imparted to the real capital stock series from using a cost index to proxy a price
index is, then, compensated for-in part? in whole? more than compensated
for?-by the fact that Philadelphia prices moved differently from national average prices, at least after 1799, and probably from 1774 as well.
There is one final problem with the deflator: it represents the costs of commercial construction in a city. A substantial fraction of the stock of structures
in the years 1799 through 1840 must have been built in the countryside by
unprofessional labor. The matter may not be very important, however. According to Adams, Philadelphia construction and Maryland farm wage rates
moved in roughly similar ways among the dates 1785, 1799, 1805,1815, and
1840.
One cannot claim great accuracy for the deflator, but on the whole it seems
satisfactory.
Animal Inventories
There are at least two problems with the animal inventory estimates. First,
they include only farm animals from 1840 onward (animals used in the mines
are part of the “equipment” estimates in mining) and probably only farm animals at earlier dates as well, whereas ideally one would like to have all domestic animals throughout. The omissions are not trivial, but neither are they
of overwhelming importance. In 1860, just over 12 percent of domestic animals, by value, were located off farms (U.S. Bureau of the Census 1860,
cviii, cxxvi, 192); in 1900, the fraction was just under 7 percent (U.S. Bureau
of the Census 1900, cxliv). The suggestion is that the total stock of animals
increased a little more slowly than did the stock of farm animals, but correcting for this shortcoming would probably not affect very substantially the conclusions previously reached.
The second problem has to do with deflation. The constant price series was
made by applying base-year prices (1860) to estimates of the numbers of animals in each year. The assumption is that a pig is a pig. In fact, pigs in 1890
were, without much doubt, superior animals to pigs of 1830. The deflator,
then, is biased, and deflation tends to understate the importance of the growth
of the stock of animals. Furthermore, the effect is also likely to be to under-
109
American Economic Growth before the Civil War
play the acceleration in the rate of growth of the per capita capital stock. The
reason is that most of the gains in the quality of animals were realized after
midcentury. In earlier decades there were probably periods when, on balance,
the quality of animal stocks actually deteriorated. Nonetheless, numbers can
reasonably proxy real values before 1840 or 1850, whereas they are less able
to perform this function thereafter. There are, of course, problems with the
evidence on numbers as well, but they seem less pressing and do not deserve
a place in this brief treatment of the subject. On the whole, the series, despite
these qualifications, is acceptable for the uses to which it has been put.
Other Inventories
The procedure followed is one employed by Kuznets (1946, 228). Inventories were taken as a fixed fraction of the value of imports and the value of
outputs of the agricultural, manufacturing, and mining sectors. No allowance
was made for changes in the efficiency with which inventories were used, a
matter of limited importance, especially before the Civil War. If there were
improvements in efficiency, then the estimating procedure tends to exaggerate
the acceleration in the rate of change of the real per capita capital stock. The
details of how the value of imports and outputs were obtained are best left to
another occasion.
Equipment
The data for the years 1840 onward were derived chiefly from the census,
were deflated by Dorothy Brady’s true price indexes, and were tested-with
considerable success-against perpetual inventory estimates (Gallman 1987).
For the earlier years, the chief sources were Jones (1980), Blodget ([1806]
1964), Goldsmith (1952), and U.S.Bureau of the Census (1975, for Treasury
data on shipping). The series seems adequate for present purposes, but should
not be trusted for much more.
Conclusions
It should be obvious that a substantial margin for error must be allowed for
all of the estimates discussed in this paper, especially those dated before 1850,
and particularly for those at the turn of the century. On the other hand, it is
not obvious that the rates of change computed from the series are subject to
important biases. The conclusions reached in sections 2.3-2.5 need not be
altered-at least not on the basis of the results of the review conducted in this
appendix.
110
Robert E. Gallman
Appendix B
Time-Quality Adjustments to the
Labor Force Estimates
This appendix describes the time-quality adjustments that were made to the
labor force estimates, for purposes of the measurement and analysis of
changes in total factor inputs and changes in total factor productivity (table
2.10). The last paragraph takes up the estimation of the elasticities of output
with respect to factor inputs that are necessary to make estimates of total factor productivity changes.
The estimates were made in two steps. First, the farm labor force figures
were adjusted to take into account changes in the farm work year.I3 Then
quality-time weights were devised for the two remaining sectors that could be
readily distinguished: mining, manufacturing, and hand trades, and all others.
The weights consisted of the ratio of labor income per worker in the relevant
sector to labor income per worker in agriculture. Since two of the important
factors accounting for sectoral differences in labor income per worker are the
relative duration of the labor year in each sector and the relative quality of
workers in each sector, one is perhaps justified in refemng to these ratios as
time-quality weights. Unfortunately, however, other factors-factors irrelevant to the time-quality adjustment-also affect intersectoral differences in
labor income per worker. Sectoral labor income deviations arose out of shortterm disequilibria in labor markets, as well as from enduring quality differences among workers. Furthermore, some part of the variations in labor income surely reflected regional and urban-rural price differences, rather than
real income disparities. It is likely that both of these factors typically operated
to widen the gaps between labor incomes in agriculture and the other two
sectors identified, each of which enjoyed higher labor incomes per worker
than did the agricultural sector. Since the labor forces attached to these two
sectors were growing faster than the agricultural labor force, the excessive
time-quality weights given these sectors mean that the rates of change of the
time-quality adjusted labor series are biased upward. The present status of
regional and urban-rural price series does not permit an appropriate deflating
of the labor income series, and there is no way of knowing how serious the
bias arising out of disequilibria in labor markets is.
There are other difficulties with these measurements.
1. It would be helpful to have detailed breakdowns of the labor force and
labor earnings, so that a more fully articulated weighting scheme might be
developed, but adequate data simply are not available.
2. Sectoral labor income estimates were developed from value-added data.
Value-added estimates involve some double-counting. If the extent of doublecounting varied from one sector to another, the labor income estimates would
13. First in principle, but not in fact. The quality adjustments were worked out first.
111
American Economic Growth before the Civil War
not be good indexes of the true relative sectoral labor incomes. It is quite
unlikely, however, that this problem is, in fact, at all serious.
3. The labor income estimates were taken as residuals, the difference between total sectoral income and sectoral property income. Property income
was estimated as the product of the value of capital and land and estimated
rates of return. Since the estimates of inventories could not be distributed
among sectors, property income was computed against the value of land and
fixed capital only. If the relative importance of inventories varied by sector,
the sectoral property estimates are biased. Unfortunately, there is no way to
be sure that this was not the case, although it is unlikely that we have here a
major source of bias.
4. More important, the system of estimating property incomes involved the
assumption that the rate of return on property ofa given rype was the same in
all sectors. In fact, this is unlikely to have been the case. The work of Bateman and Weiss shows that the returns to property in the antebellum South
were much higher in manufacturing than in agriculture (1981, 107, 108,
114).l4Unfortunately, there is no good basis for producing different sectoral
rates of return for all types of property for all sectors in all years. We can be
quite sure, however, that the procedure followed to produce labor income estimates has led to an exaggeration of the relative levels of labor income in the
“mining, manufacturing, and hand trades” sector, and probably in the “all
other” sector as well. This in turn means that the time-quality weights attached to the nonfarm sector labor forces are too high and that, therefore, the
rates of change of the adjusted labor series are biased upward.
The sectoral value-added series (current prices) were taken from volumes
24 and 34 of Studies in Income and Wealth (Gallman 1960, 47, 54, 56, 63;
Gallman and Weiss 1969, 305), and were adjusted in the following ways. The
estimates of farmland improvements were dropped from farm value-added,
and new estimates, derived from data in volume 30 of Studies in Income and
Wealth (Brady 1966) were substituted for them.I5 Value added by the “all
14. The rates of return I have used do vary from one sector to another, as the structure of the
capital stock varies; only the rates for individual types of property are constant. But the differences
in the average rates that have emerged are small, compared with the ones observed by Bateman
and Weiss. For example, the average rates I have obtained in four of the years are
Agriculture
Manufacturing, mining, and hand trades
All other
1840
1860
1880
1900
1 1.6%
13.0
13.2
11.O%
12.6
12.5
9.4%
10.9
10.7
7.6%
9.4
8.9
Bateman and Weiss (1981, 116) report rates of return for large manufacturing firms of 17% in
1850, and 21% in 1860.
15. Gallman 1966, 35, variant I. The estimates are available in constant prices only. Current
price estimates were made by assuming that the ratio of improvements to farm value added was
the same in current and constant prices. The average value of improvements for 1834-43 was
taken to correspond to the value of improvements in census year 1839, and so forth. The ratio of
the value of improvements to the value of farm value added in 1859 was estimated on the basis of
112
Robert E. Gallman
other” sector was formed by adding to total value added by services (taken
from volume 34 of Studies in Income and Wealth), value added by construction (drawn from volume 24 of Studies in Income and Wealth, construction
variant A), and then subtracting the value of shelter and value added by the
hand trades. The value of shelter was dropped because the production of shelter involves the use of practically no labor and therefore the value of shelter
should not figure in the estimation of sectoral labor quality weights. Value
added by the hand trades was added to value added by manufacturing and
mining, taken from volume 24.
The gross rate of return for each type of property is composed of the net
rate plus depreciation (if any). The following depreciation rates were assumed: Land, 0; animals, 0; buildings, fences, irrigation, and drainage works,
2 percent; land clearing and breaking, 0; tools and equipment, 6.67 percent.
The net rate of return was taken to be 10 percent in 1860 and was adjusted in
the other years on the basis of an index number of the rate of return on New
England municipal bonds (Homer 1963, 287-288, linked at 1857-59 to Boston City 5s, 305).
The labor force data were drawn from Weiss’s paper in this volume. The
division of the nonfarm labor force between the two nonfarm sectors was
based on Lebergott ( 1964).
The adjustment for changes in agricultural work hours was based on data in
Gallman (1975, 73), and the David, Lebergott, and Weiss series. From Gallman (1975, 73, inclusive of improvements, variant B), and the David and
Lebergott farm labor force series, it was possible to compute an index of the
hours worked by farm laborers in 1800, 1850, and 1900. With this index and
the Weiss farm labor force in each of these three years, an index of the number
of hours worked per worker was computed. Index numbers for the missing
intermediate years were interpolated on a straight line. The index for 1774
was assumed to be the same as the index for 1800. The aggregate qualityadjusted labor force series were then adjusted for changes in the number of
hours worked by multiplying them by the index of hours worked per worker.
The procedure adopted to make estimates of the elasticities of output with
respect to inputs was similar to the one by which labor and property incomes
were computed for the three sectors (see above). The only difference was that
the calculations were made at the national, not the sectoral, level and that
components of capital left out of the sectoral calculations-inventories, the
international sector-were here added back in.
the ratio of improvements, 1849-58, and farm value added 1854. A similar procedure was followed to obtain the ratio for census year 1869.
113
American Economic Growth before the Civil War
References
Abramovitz, Moses. 1989. The nature and significance of Kuznets cycles. In Thinking
about growth and other essays on economic growth and welfare, 245-75. Cambridge: Cambridge University Press.
Adams, Donald R. 1975. Residential construction industry in the early nineteenth century. Journal of Economic History 35: 794-816.
. 1986. Prices and wages in Maryland, 1750-1850. Journal of Economic History 46, no. 3: 625-45.
Batemen, Fred, and Thomas Weiss. 1981. A deplorable scarcity. Chapel Hill: University of North Carolina Press:
Bezanson, Anne, Robert D. Gray, and Miriam Hussey. 1936. Wholesale prices in
Philadelphia, 1784-1861: Part I . Industrial Research Study no. 29. Philadelphia:
University of Pennsylvania. Reprinted in U.S. Bureau of the Census, Historical
statistics of the United States, ser. E-97, E-111. Washington, D.C.: GPO,1975.
Bishir, Catherine W., Charlotte V. Brown, Carl R. Lounsbury, and Ernest H. Wood,
111. 1990. Architects and builders in North Carolina: A history of the practice of
building. Chapel Hill: University of North Carolina Press.
Blodget, Samuel. [1806] 1964. Economica: A Statistical manual for the United States
of America. Reprint. New York: Augustus Kelley.
Brady, Dorothy S. 1966. Price deflators for final product estimates. In Output, employment, and productivity in the United States afer 1800, ed. Dorothy s. Brady, 91115. NBER Studies in Income and Wealth, 30. New York Columbia University
Press.
David, Paul, and Peter Solar. 1977. A bicentenary contribution to the history of the
cost of living in America. Research in Economic History 2: 1-80.
Davis, Lance E., Richard A. Easterlin, William N. Parker, et al. 1972. American
economic growth: An economist’s history of the United States. New York: Harper
and Row.
Davis, Lance E., and Robert E. Gallman. 1978. Capital formation in the United States
during the nineteenth century. In The Cambridge economic history of Europe, vol.
7, pt. 2: 1-69,496-503,557-61. Cambridge: Cambridge University Press.
Engerman, Stanley L., and Robert E. Gallman. 1983. U.S. economic growth, 17901860. Research in Economic History 8: 1-46.
Fishlow, Albert. 1966a. The American common school revival: Fact or fancy? In Industrialization in two systems, ed. Henry Rosovsky, 40-67. New York John Wiley
and Sons.
. 1966b. Levels of nineteenth century American investment in education. Journal ofEconomic History 26: 418-36.
Fogel, Robert W., and Stanley L. Engerman. 1977. Explaining the relative efficiency
of slave agriculture in the antebellum south. American Economic Review 67:
275-96.
Gallman, Robert E. 1960. Commodity output, 1839-1899. In Trends in the American
economy in the nineteenth century, ed. William N. Parker, 13-67. NBER Studies in
Income and Wealth, 24. Princeton: Princeton University Press.
. 1966. Gross national product in the United States, 1834-1909. In Output,
employment, andproductivity in the United States afer 1800, ed. Dorothy S. Brady,
3-76. NBER Studies in Income and Wealth, 30. New York: Columbia University
Press.
. 1972. Changes in total U S . agricultural factor productivity in the nineteenth
century. Agricultural History 46: 19 1-209.
114
Robert E. Gallman
. 1975. The agricultural sector and the pace of economic growth: U.S. experience in the nineteenth century. In Essays in nineteenth century economic history, ed.
David C. Klingaman and Richard K. Vedder, 35-76. Athens: Ohio University Press.
. 1986. The United States capital stock in the nineteenth century. In Long-term
factors in American economic growth, ed. Stanley L. Engerman and Robert E. Gallman, 165-206. NBER Studies in Income and Wealth, 51. Chicago: University of
Chicago Press.
. 1987. Investment flows and capital stocks: U.S. experience in the nineteenth
century. In Quantity and quiddity: Essays in US.economic history in honor of Stanley L. Lebergott, ed. Peter Kilby, 214-54. Middletown, Conn.: Wesleyan University Press.
Gallman, Robert E., and Thomas Weiss. 1969. The service industries in the nineteenth
century. In Production and productivity in the service industries, ed. Victor R.
Fuchs, 287-381. NBER Studies in Income and Wealth, 34. New York: National
Bureau of Economic Research.
Giffen, Robert. 1889. Growth of capital. London: George Bell and Sons.
Goldin, Claudia, and Robert A. Margo. 1989. Wages, prices, and labor markets before
the civil war. NBER Working Paper no. 3198. Cambridge, Mass.: National Bureau
of Economic Research.
Goldsmith, Raymond W. 1952. The growth of the reproducible wealth of the United
States of America, 1805 to 1905. In Income and wealth of the United States: Trends
and structure, ed. Simon Kuznets, 247-328. Income and Wealth Series, 2. Cambridge: Bowes and Bowes.
. 1982. The national balance sheet of the United States, 1953-1980. Chicago:
University of Chicago Press.
. 1985. Comparative national balance sheets: A study of twenty countries,
1688-1978. Chicago: University of Chicago Press.
Homer, Sidney. 1963. A history of interest rates. New Brunswick, N.J.: Rutgers University Press.
Jones, Alice Hanson. 1980. Wealth of a nation to be. New York: Columbia University
Press.
Kuznets, Simon. 1938. On the measurement of national wealth. In Studies in income
and wealth, 2, 3-61. New York: National Bureau of Economic Research.
. 1946. Nationalproduct since 1869. New York: National Bureau of Economic
Research.
Lebergott, Stanley L. 1964. Manpower in economic growth. New York: McGraw-Hill.
. 1985. The demand for land: The United States, 1820-1860. Journal of Economic History 65: 18 1-2 12.
Pitkin, Timothy. 1835. A statistical view of the commerce of the United States of America. Reprint. Greenwich, Conn.: Johnson Reprint Corporation, 1967.
Primack, Martin. 1962. Farm formed capital in American agriculture, 1850 to 1910.
Ph.D. diss., University of North Carolina, Chapel Hill.
Rothenberg, Winifred. 1988. The emergence of farm labor markets and the transformation of the rural economy: Massachusetts, 1750-1855. Journal of Economic History 48: 537-66.
Seaman, Ezra C. 1852. Essays on the progress of nations. New York: Charles
Scribner.
Smith, Walter B. 1963. Wage rates on the Erie Canal. Journal of Economic History
23, no. 3: 298-311.
Soltow, Lee. 1984. Wealth inequality in the United States in 1798 and 1860. Review of
Economics and Statistics 66: 444-5 1 .
115
American Economic Growth before the Civil War
U.S. Bureau of the Census. 1860. Agriculture of the United States in 1860. Washington, D.C.: GPO.
. 1900. Agriculture: Part 1: Farms, live stock, and animal products. Washington, D.C.: GPO.
. 1975. Historical statistics of the United States, colonial times to 1870: Bicentennial edition: Parts 1 and 2 . Washington, D.C.: GPO.
Uselding, Paul. 1971. Conjectural estimates of gross human capital inflows to the
American economy, 1790-1860. Explorations in Economic History 9: 49-62.
Warren, George F., and Frank A. Pearson. 1933. Prices. New York: John Wiley and
Sons. Reprinted in US. Bureau of the Census, Historical statistics of the United
States, ser. E-52, E-59. Washington, D.C.: GPO, 1975.
Wheeler, Gervase. 1855. Homes f o r the people in suburb and country. New York:
Charles Scribner.
Comment
Stanley L. Engerman
Robert Gallman here uses measures of the capital stock to estimate and describe the pattern of economic growth in the United States from the Revolutionary period to the end of the nineteenth century. The measures represent a
continuation of his ongoing work, previously published in several places.’ In
these earlier publications he has presented many of the details of calculation
for the 1840 to 1900 estimates, as well as described the various concepts and
tests going into their preparation. In general, most of these problems are wellknown and ably discussed, so there can be little new to say here in regard to
the major issues. Following another remark of Giffen’s, we can only compliment Gallman for doing the best that can be done with the limited data available, and though the “figures are necessarily rough,” they make “a little clear
what would otherwise be most dark, and they suggest problems for inquiry
which would not otherwise be thought of.” For “the figures, though rough,
can be reasoned on safely with care.”2
There is one initial point about the basic concept of capital that Gallman
uses that is worth noting. His measures are restricted to variants of physical
capital. There are no estimates of human capital, even of the slave population
for which market values do exist. But, compared to the familiar constructs of
Goldsmith, Gallman’s measure is not of all tangible wealth, since he does not
Stanley L. Engerman is John H. Munro Professor of Economics and professor of history at the
University of Rochester and a research associate of the National Bureau of Economic Research.
1. Robert E. Gallman, “The United States capital stock in the nineteenth century,” in Long-term
factors in American economic growth, ed. Stanley L. Engerman and Robert E. Gallman, NBER
Studies in Income and Wealth, 51 (Chicago: University of Chicago Press, 1986), 165-206; and
“Investment flows and capital stocks: U.S. experience in the nineteenth century,” in Quantity and
quiddity: Essays in U.S. economic history in honor of Stanley L. Lebergott, ed. Peter Kilby
(Middletown, Conn.: Wesleyan University Press, 1987), 214-54.
2. Robert Giffen, Growth ofcapital (London: George Bell and Sons, 1889), 157.
116
Robert E. Gallman
include all of the value of land, nor is it all reproducible tangible wealth, since
he does include some of the value of land.3 Rather, Gallman’s estimates include the value of improvements made to land, approximately equal in some
years, it turns out, to the market value of improved land. Although Gallman’s
measure does omit the value of privately owned unimproved land and the
“pure rent” on the acres improved, in most years the value of improved land
represents the largest part of the total value of land. Thus the distinction between the Goldsmith and the Gallman treatments of land, while interesting
and important, will not seriously distort most long-term comparisons.
Gallman uses the capital stock both as a means of measuring economic
growth, for a period of time for which the basis of income measurement is not
readily available, and as part of the explanation of the nature of economic
growth, using capital stock measures with related input data to describe the
patterns of change. There are some points to consider in the use of changes in
the capital stock to measure changes in income level, as well as some differences between measures of potential income (which is perhaps the most desired measure of economic growth), observed (measured) income, and capital. There are choices made out of potential income that influence observed
income and the observed capital stock. The choices between goods and leisure, and decisions in regard to fertility, clearly influence measured output per
capita, as do the effects of intensity avoidance, risk avoidance, and market
avoidance upon the product mix and thus potential income forgone. Capital,
being based upon the amounts of income not consumed in the past, will have
a different growth rate than income if there are changes in the savings rate
over time. In addition, in considering the effects of savings upon the capital
stock and its measured potential for future growth, it is also necessary to consider the form that these savings and investments take, and the related differences in types of assets and in their longevity. Savings can be used to provide
either producer durables or consumer durables (the latter are omitted by Gallman, except for dwellings). It has, for example, been argued that in English
history the level of savings was long sufficient to have financed the industrial
revolution, but that a change in its structure and composition was needed for
long-term g r ~ w t h Further,
.~
as Gallman points out, the longevity of capital
will influence the breakdown between gross and net investment, and of the
available capital stock. Estimates of asset durability and obsolescence are not,
as Gallman notes, independent of the performance of the economy and its rate
3. See, for example, Raymond W. Goldsmith, “The growth of the reproducible wealth of the
United States of America, 1805 to 1905,” in Income and wealth of the United States: Trends and
structure, ed. Simon Kuznets, Income and Wealth Series, 2 (Cambridge: Bowes and Bowes,
1952), 247-328; and The national balance sheer of the United States, 1953-1980 (Chicago: University of Chicago Press, 1982).
4. M. M. Postan, “Recent trends in the accumulation of capital,” Economic History Review 6
(1935): 1-12.
117
American Economic Growth before the Civil War
of technical change. Kuznets suggests that one of the basic shifts from the
premodern to the modem era was not in the gross investment rate, but rather
reflected an increase in asset longevity, lowering capital consumption, and
leading to a larger capital stock from the gross in~estment.~
How good a proxy for the rate of growth of income is the rate of growth of
the capital stock? In the long-run, increases in both are part of modem economic growth, and high growth (relative to preindustrial times) of both will
come together. In the United States, the capital stock generally grew more
rapidly than did GNP throughout the nineteenth century (and, depending on
which of the several variants used from table 2.4, possibly after 1774), accounting for the rising capital-output ratio over that period. For shorter intervals, however, there are also differences not only in magnitudes, but even in
the comparative ranking of rates of change, and the periods of acceleration or
deceleration can differ. Thus the choice between income and capital as a measure of growth can influence examinations of the turning points in the growth
process. Nevertheless, since for the period before 1840 it seems more possible
to build up capital estimates from probate inventories and tax reports than to
generate income data when no census production (or labor force) data are
available, clearly these capital stock estimates for the early period must provide an essential set of measurements to be used by economic historians in the
quantitative study of economic growth.
Gallman’s estimates indicate that the post- 1840 years of the nineteenth century had rates of growth of the capital stock per capita more rapid than those
in preceding years.6 In particular, the 1840 to 1860 growth rate exceeded that
of the pre-1840 years and, indeed, almost all twenty-year periods since. This
is not the same as dating the acceleration of growth in 1840, since the most
analyzed data are for 1774, 1805, and 1840 (with 1815 given less attention).
Thus it is hard to pinpoint from the data exactly when after 1805 (or 1815) the
growth spurt began, but clearly some increase in capital’s growth rate occurred in the first half of the nineteenth century (for at least three of the four
series shown in table 2.6). For the last quarter of the eighteenth century the
capital stock probably also grew at higher rates than did income, particularly
for the concept of capital that excludes land clearing. Note, also, that the
1774-1840 rate of increase of the U.S. capital stock was considerably above
that for Great Britain in this period.’
5 . Simon Kuznets, “Capital formation in modem economic growth (and some implications for
the past),” in Third International Conference of Economic History (Paris: Mouton, 1968).
6. In general, unless otherwise stated, the comparisons will be based on the Gallman capital
stock series including land clearing and breaking, with the consumer price index as a deflator.
While most comparisons would not differ if any of the other series were used, some would require
minor alteration.
7. C. H. Feinstein, “Capital formation in Great Britain,” in The Cambridge economic history
ofEurope (Cambridge: Cambridge University Press, 1978). vol. 7, pt. 1.
118
Robert E. Gallman
The growth pattern of the U.S. capital stock was consistent with the general
pattern of growth in income, which also had an acceleration in the interval
between 1800 and 1840. Compared to Thomas Weiss’s (chap. 1 in this volume) newest estimates of national income, and using the estimated capital
stock including land clearing, capital grew at a more rapid rate than did income between 1800 and 1840, while both had considerably higher rates of
growth after 1840 than before, with the shift upward in the growth of capital
being sharper than that in income. Weiss has some acceleration in the growth
of per capita income after 1820, consistent with Gallman’s post-1815 acceleration in capital stock growth.8
Significant structural shifts in the composition of the capital stock occurred,
particularly when looking at the measure of capital stock including the value
of land improvements. The relative shares of the other four major components
of capital changed relatively little prior to 1840, particularly in constant dollar
measures. The share of land improvements declined sharply starting with the
1774 estimates, while a quite dramatic decline in the share of animal inventories began in 1840. (There was possibly a smaller reduction in the food obtained from these inventories.) Both declines reflect the relative reduction in
the role of the agricultural sector in the economy. There was a sharp rise in the
share of equipment in constant dollar estimates (influenced by the fall in the
relative price of equipment compared to construction), which starts around
1840 and accelerates after 1870.
As noted, one cause of the shift in the structure of capital was the decline
in the share of the agricultural sector in the nineteenth century, a decline that
in the Weiss labor force estimates was particularly sharp in the period from
1840 to 1860. There was a considerably greater decline in the capital share in
agriculture than in Weiss’s labor force estimates for 1800 to 1860, and overall
a larger, but smoother, decline after 1840 than that in the Lebergott labor force
series. And, if we use equipment as a rough measure of modernization, the
fact that in 1840 it accounted for only 6 or 8 percent of the constant dollar
capital stock suggests that up to that time increased investment in this component had only a limited potential for influencing the overall measured rate
of growth of the economy, a point also indicated by the estimates of the share
of the labor force in manufacturing.
It will be useful to place some of the issues raised by Gallman’s capital
stock estimates in a broader international and intertemporal perspective, particularly in comparisons with the other major developing economy of the late
eighteenth and early nineteenth century, Great Britain.
First, the United States had high growth rates in total and per capita capital
stock from quite early years, rates of growth not achieved by many other
countries until the period after World War 11, a finding rather similar to that
8. Note that Margo (chap. 4 in this volume), whose series for real wages begins in 1820, finds
a shift upward in real wages in the 1830s or (in his preferred series) the 1840s.
119
American Economic Growth before the Civil War
for the growth of i n ~ o m eIndeed,
.~
U.S. capital stock, total and per capita,
grew much more rapidly than did that for the “first industrial nation,” Great
Britain, between the late eighteenth century and the end of the nineteenth.
Second, the high U.S. capital stock growth was accomplished with, before
1840, a savings ratio rather low by later standards.’O The U.S. ratio of savings
to income was probably lower than that of the British in the years between
1760 and 1840, the British ratio rising somewhat earlier (between 1760 and
1800) than that of the United States. After 1840 the U.S. ratio rises, sharply
after the Civil War, and remains high through the remainder of the nineteenth
century, Unlike the United States savings ratio the British ratio remained basically unchanged throughout the nineteenth century. The United States had,
by the late nineteenth century, the highest savings rate among developed countries, rates not reached by many countries until after World War 11.
Third, while the United States had a slightly higher share of land in total
tangible wealth at the start of the nineteenth century than did Great Britain, in
subsequent years of the nineteenth century both countries had roughly similar
declines in the land share, although there was significant decade-to-decade
variability. What might seem noteworthy, given the major differences in geographic expanse, industrial structure, and so forth, is the relative smallness of
the intercountry differences in the shares of land in total wealth.
Fourth, the United States had a considerably lower capital-output ratio than
did Great Britain (and the rest of the world) in the nineteenth century. Of the
twenty countries for which Goldsmith provides data on capital-output ratios
for net tangible assets, no country before 1939 had a capital-output ratio as
low as that of the United States in 1850, with the exception of India.” The
United States had, at the onset of growth, probably the lowest average ratio of
capital to output of all developing countries; the British ratio of tangible
wealth to income in 1800 was several times that of the United States. Post1840, however, there were sharp rises in the United States in both the rate of
savings out of current income and the ratio of capital to output. By the end of
the century the U.S. capital-output ratio was among the lower, but was by no
means the lowest, among developed countries. For the British, the capitaloutput ratio fell sharply throughout the century. The relative differences in the
movement of output growth per unit of capital growth pose some interesting
comparative questions for studies of the sources of growth.
Gallman’s basic findings regarding the growth of the capital stock and its
acceleration in the first part of the nineteenth century present a pattern similar
to that of the growth of income, and all of this seems quite plausible given
9. See Raymond W. Goldsmith, Comparative national balance sheets: A study of twenty counrries, 1688-1978 (Chicago: University of Chicago Press, 1985).
10. See Lance E. Davis and Robert E. Gallman, “Capital formation in the United States during
the nineteenth century,” in The Cambridge economic history of Europe (Cambridge: Cambridge
University Press, 1978), vol. 7 , pt. 2. Gallman, “United States capital stock.”
1 1 . Goldsmith, Comparative national balance sheets.
120
Robert E. Gallman
other sources and the present state of knowledge. Thus the consistency is
somewhat reassuring.
These capital stock estimates for the early national period pose many familiar historical questions, which Gallman has discussed here and elsewhere.I*
How were they financed? What individual behavior led to this new level of
savings and investment? Did this require the formation of new institutions and
legal provisions? And, given the role of land clearing and its importance, how
much did it cost in terms of forgone output or was it, for the most part, forgone leisure? And, if the latter, was this increase in investment in improvements undertaken due to a taste change in favor of goods or was it due to a
shift in the opportunity cost of time? Thus, as Gallman suggests, what seem
for some purposes to be interesting questions for measurement are also significant issues for the broad understanding of the historical process.
12. As have Davis and Gallman in “Capital formation in the United States.”